Research topics: Differential Privacy, Neural Network Verification
Publications (1st author)
MpBP: Verifying Robustness of Neural Networks with Multi-path Bound Propagation
Code
Slides
Slides-2
Ye Zheng, Jiaxiang Liu, Xiaomu Shi
ESEC/FSE Demonstration 2022: European Software Engineering Conference and Symposium on the Foundations of Software Engineering
基于多路径回溯的神经网络验证方法
and the English version
Multi-path Back-propagation Method for Neural Network Verification
Code
Slides
郑烨，施晓牧，刘嘉祥
JOS 2022 and IJSI 2022: 软件学报 and International Journal of Software and Informatics
Archives
硕士学位论文《多路径方法在神经网络验证中的研究与应用》
PDF
Slides
本文关注神经网络验证方法中界限传播方法的精度问题。关于此问题，本文提出界限传播路径的概念，将各种界限传播方法扩展到其对应的多路径界限传播方法；此外，本文将多路径界限传播在 PyTorch 框架上并行化，开发了高效而易用的鲁棒性验证工具。
More Bio
Shenzhen University (2020 - 2023)
- Master's degree in Software Engineering
- Research areas: Neural Network Verification and Formal Verification, advised by Dr. Jiaxiang Liu
Henan University (2016 - 2020)
- Bachelor's degree in Mathematics
- Majored in Pure Mathematics, advised by Dr. Zhonghua Wang